INDEX

Explanations

comma and

This neuron fires most strongly on the key content word at the end of each abstract or summary—i.e. the single noun or verb that names the document’s main object or action.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

rfan

-1.88

 lorsque

-1.87

 テープ

-1.85

modium

-1.77

 quando

-1.73

 diante

-1.73

鶫

-1.72

its

-1.67



-1.66

 tetrachloride

-1.65

POSITIVE LOGITS

and

4.56

has

2.08

had

2.05

at

1.88

 gave

1.80

 became

1.80

There

1.80

an

1.79

 have

1.77

1.75

Activations Density 0.149%